Update app.py

app.py

@@ -42,7 +42,7 @@ print(f"Loading model: {model_name} with quantization")
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
     quantization_config=quantization_config_4bit, # Comment out if not using quantization
-    device_map="auto",
+    device_map="cuda",
 )
 
 print(f"Loading tokenizer: {model_name}")